How DeployGPU pricing works

The price on a DeployGPU offer is the price you pay per hour: the rate is set by the provider, any platform fee is already inside it, and the deploy page breaks it down line by line before you launch. You pay from prepaid credits, so the balance you add is also the most you can spend. Everything else on this page is detail about how those hours are counted and refunded.

The short version

• You pay the hourly rate shown on each offer, for the time the instance runs.
• You add credits in advance by card, from $5. Each top-up carries a card processing fee of 5% + $0.30.
• Launching reserves the first hour. When you stop, the unused seconds of the current hour go back to your balance.
• A launch that fails, or one you stop before it is Running, is refunded in full, automatically.
• No contracts, no minimum term, no card needed to sign up.

Starting prices right now

Every price below is a VM you can launch now, cheapest offer per GPU, refreshed every 15 minutes. The GPU pages list every offer with its region, CPU, RAM and disk.

What is inside the hourly rate

The hourly rate on an offer covers the GPU machine from the provider, the disk that comes with it and, where a provider charges for them, extra disk and a public IP address. Offer prices include any platform fee. The deploy page shows the full breakdown before you pay, rounded to the cent, and that rounded rate is exactly what you are billed.

Disk is fixed on most offers. On Verda you can choose a larger disk at launch, up to 500 GB where the offer allows it, and the extra gigabytes are added to the hourly rate you see before you confirm.

Adding credits

Credits are bought in US dollars through Stripe, by card or with Stripe Link, from $5 to $25,000 per top-up. Every top-up, including charges to a saved card and automatic top-ups, adds a processing fee of 5% + $0.30, shown as its own line at checkout. Your balance grows by the amount you chose, and the fee is charged on top:

Because the $0.30 part is fixed, fewer and larger top-ups cost less in fees than many small ones. Auto top-up, on the billing page, charges your saved card when the balance drops below a threshold you set between $2 and $1,000, at most once an hour.

Credit purchases are final under the Terms of Service: unused credit stays in your account for the next launch rather than going back to the card.

How the hours are counted

Billing runs on prepaid hourly blocks settled to the second:

1. When you launch, the first hour is reserved from your balance. Your balance must cover that hour for the launch to start.
2. Paid time starts when provisioning starts. If the instance reaches Running, those provisioning minutes are part of what you pay for.
3. Each new hour is charged when it begins.
4. When you stop, the unused seconds of the current hour are refunded to your balance.

Two cases cost nothing. If a launch fails on the provider's side, the reserved hour is refunded automatically. If you stop an instance before it reaches Running, the whole reserved hour comes back too. And if a provider loses a running machine, the health check that runs every 5 minutes stops the deployment and refunds the unused time.

Worked examples

These use an example rate of $1.00 per hour so the arithmetic is easy to follow. Real rates are on each offer.

The processing fee sits outside these numbers: it is paid once, when you add credits.

When your balance runs low

When your balance cannot cover the next full hour, the instance does not stop at the hour mark. It switches to per-minute billing and keeps running until the balance reaches zero. Then it is stopped, and like any stop, its disk is deleted.

Emails go out along the way: one when about 2 hours of runtime are left, one when per-minute billing starts, and one when the instance is stopped, plus a low-balance email whenever a charge leaves the balance under $2. Auto top-up, when it is on, runs before any of them. There is no grace period after the balance reaches zero, which is also why prepaid credits work as a spending cap: without auto top-up, you cannot spend more than you added.

Storage

Instance disk is part of the hourly rate and is deleted when the instance stops. For data that must outlive an instance, Verda and Hyperstack offer volumes. A volume is billed every hour it exists, attached or not, at the provider's storage price, which the dashboard shows when you create one. On 28 September 2026 that worked out to about $0.20 per GB per month for Verda NVMe volumes and about $0.07 per GB per month for Hyperstack volumes.

The stop guide covers how to keep data between sessions.

Other charges and things that are not sold

Vast.ai offers, which you can only launch from the dashboard, also bill network bandwidth by usage, at Vast.ai's rate plus 10%. Today that is the only usage charge outside the hourly rate.

A few things are not for sale on DeployGPU: spot or interruptible instances, reserved pricing you can buy yourself, and invoices. Receipts arrive by email for every payment, and the transaction history on the billing page exports to CSV. Clusters and multi-month terms are quoted by a person through the cluster request form.

The offer prices come from the same scrapers that feed GpuPerHour, and every provider DeployGPU launches on is checked every minute. For prices across the rest of the market, including providers you cannot launch on here, compare all providers on GpuPerHour.